Analysis Report — telecom

Generated on 2025-10-06T21:39:42.048683
Rows: 800
Positive: 34.1% • Negative: 41.4% • Neutral: 24.5

Executive summary

Overall sentiment: mean VADER compound = -0.071 (std 0.305).
Top positive terms: service, charge, speak, need, mobile, suddenly
Top negative terms: dropped, area, causing, frequent, frustrating, unstable

Notable time points

Most positive period: 2025-08-27 (mean=-0.005).
Most negative period: 2025-09-24 (mean=-0.11).

Top Terms

Sentiment Over Time

Predictive terms (predicting negative sentiment)

The table below lists features (tokens) that are most predictive of the negative label based on a simple logistic model.
term coef
area 2.936128
causing 2.906075
dropped 2.802943
frequent 2.797740
ref 1.642532
unstable 1.602901
frustrating 1.034478
umm 0.947677
outage 0.854577
know 0.846228

Download CSV

Sentiment spikes

Detected spikes in rolling sentiment (z-score threshold). Review the rows to understand dates and severity.
timestamp sentiment_mean count rolling_mean rolling_std zscore spike
2025-07-09 -0.101766 58 -0.101766 0.000000 0.000000 False
2025-07-16 -0.082558 66 -0.092162 0.013582 0.707107 False
2025-07-23 -0.083025 60 -0.089116 0.010957 0.555889 False
2025-07-30 -0.060942 60 -0.075508 0.012617 1.154502 False
2025-08-06 -0.067121 63 -0.070362 0.011393 0.284543 False
2025-08-13 -0.064564 64 -0.064209 0.003105 -0.114429 False
2025-08-20 -0.047902 62 -0.059862 0.010437 1.146007 False
2025-08-27 -0.004697 60 -0.039054 0.030899 1.111935 False
2025-09-03 -0.047007 61 -0.033202 0.024690 -0.559130 False
2025-09-10 -0.100200 64 -0.050634 0.047855 -1.035747 False

Download CSV

Topic — sentiment correlation

Correlation between LDA topic weights and VADER sentiment (positive values indicate topics associated with more positive sentiment).
topic keywords corr_with_sentiment
topic_0 overpriced, dropped, causing, frequent, area, outage 0.091041
topic_1 dropped, area, causing, frequent, frustrating, intermittent -0.408279
topic_2 service, charge, speak, need, warning, suddenly 0.522061
topic_3 hour, past, unlimited, helpful, intermittent, coverage 0.206398
topic_4 dropped, area, frequent, causing, unstable, poor -0.487181
topic_5 incorrect, recent, invoice, resolved, data, plan 0.143191

Download CSV

Clusters — top terms & samples

Clusters are produced by LSA embeddings + KMeans; top terms summarize each cluster and sample snippets give context.

Cluster top terms

Cluster 0: speak, charge, need, excellent, unlimited, customer, invoice, intermittent, helpful, billing

Cluster 1: coverage, frequent, causing, area, dropped, overpriced, unstable, frustrating, poor, ref

Cluster 2: check, location, service, card, sim, plan, data, network, coverage, signal

Cluster 3: warning, mobile, suddenly, recent, incorrect, resolved, upgraded, restarted, disconnected, escalated

Cluster 4: dropped, frequent, causing, area, frustrating, outage, overpriced, poor, unstable, intermittent

Cluster 5: hour, past, unlimited, helpful, intermittent, excellent, coverage, card, sim, invoice

Cluster samples

Cluster 0:
need speak outage overpriced charge
need speak billing helpful charge
need speak sim card excellent charge

Cluster 1:
unstable coverage area causing frequent dropped call
overpriced coverage area causing frequent dropped call
overpriced coverage area causing frequent dropped call

Cluster 2:
check service signal location
check umm service sim card location
check service invoice location

Cluster 3:
restarted incorrect amount recent customer service
escalated incorrect amount recent invoice
restarted incorrect amount recent invoice

Cluster 4:
unstable signal area causing frequent dropped call
poor signal area causing frequent dropped call
poor data plan area causing frequent dropped call umm

Cluster 5:
customer service unstable past hour
customer service unlimited past hour
roaming intermittent past hour

Download cluster top terms • Download cluster samples